Research On Key Issues For Development Of SLAM Algorithm Of Explosion-proof Inspection Robots

SLAM technology is one of the key technologies to realize intelligent applications for inspection robots.This study takes the self-developed explosion-proof inspection robot as the research object,which is charactered with four-wheel independent drive and independent steering.It is equipped with lidar and inertial sensors as the environment perception and positioning sensors,and builds it on the ROS platform.A complete set of SLAM system framework to realize the autonomous positioning and map construction functions of the explosion-proof inspection robot in the natural gas pressure regulating station.First of all,according to the natural gas pressure regulating station operating environment and explosion-proof requirements,a four-wheel independent drive and independent steering explosion-proof inspection robot is designed with strong adaptability and high safety.Aiming at the existing explosion-proof technology,battery,industrial computer,IMU and other electrical equipment is adopted for the overall explosion-proof designed.Considering that the lidar needs to be installed outside the car body and its performance requirements,a separate explosion-proof design is carried out to make the overall explosion-proof requirement of the robot reach Ex de IIB.In addition,the structural optimization design of the explosion-proof enclosure is carried out according to the characteristics of the stiffener,so that the robot has good load-bearing performance and compact mechanical structure,and the mass of the explosion-proof enclosure can be reduced by 19.2% under the premise of meeting the requirements of explosion-proof and strength.Afterwards,in view of the problems that the explosion-proof drive motor of the robot has no encoder and can’t effectively establish the kinematics model of the robot wheel odometer,therefore this paper can quickly and accurately estimate the plane motion of the lidar based on the two frames of data within the continuous scanning range of the lidar,with establishment the motion model of the laser odometer.Further,in order to make the laser odometer better applicable to the dynamic unstructured environment,computer vision optical flow technology is used for reference,and the symmetrical range flow constraint model of the two frames of laser data is built to obtain a more accurate Motion estimation.In addition,the laser odometer drift problem is easily caused during the pivot steering movement of the robot.The EKF algorithm is used to fuse the laser odometer and IMU sensor data effectively improves the drift problem of the robot when turning.Finally,the SLAM algorithm of the four-wheel independent driving and independent steering robot is researched.Based on the use of laser odometer positioning,by studying the RBPF algorithm based on particle filtering,the Karto and cartographer algorithms based on graph optimization,Aiming at the shortcomings of traditional algorithms,this study proposes an improved RBPF algorithm based on the traditional adaptive resampling technology,and proposes a segmented multi-threshold importance resampling technique.The simulation and experimental investigations show that the improved algorithm can effectively increase particles diversity and mitigate particle degradation problems,and the positioning error can be reduced by 20%.